strongwolf/DW

Issues

• Why multiple 0.25?

  #42 opened 8 months ago by yuanqianguang
  0

• Porting DW to target tracking

  #41 opened a year ago by Jontyshaw
  0

• The difference beteen Autoassign and this DW

  #33 opened 2 years ago by CPFelix
  3

• How to apply DW to two-stage detectors？

  #40 opened a year ago by seacloud-0420
  1

• Are mean and sigma in CenterPrior learnable?

  #39 opened a year ago by Yuanyang-Zhu
  1

• weight design

  #36 opened 2 years ago by gnftau
  3

• About the version of mmcv

  #38 opened 2 years ago by liangabcd
  1

• oom问题

  #37 opened 2 years ago by Yangjie0610
  0

• how to understand this

  #35 opened 2 years ago by sqdcbbj
  0

• Welcome update to OpenMMLab 2.0

  #34 opened 2 years ago by vansin
  0

• At inference, a neg prediction in the ranking list will not affect the recall but decrease the precision.

  #31 opened 2 years ago by ycyinchao
  1

• Why use neg_avg_factor = (1 - p_neg_weight).sum() ?

  #30 opened 2 years ago by Icecream-blue-sky
  1

• About bbox refinement

  #32 opened 2 years ago by satuoqaq
  2

• Objectness

  #29 opened 2 years ago by nqxjzm
  2

• Expect dw to bring more gain in yolov6！

  #16 opened 2 years ago by meituan-gengyifei
  2

• I have a question about wneg=1-wpos

  #28 opened 2 years ago by ycyinchao
  0

• 请问可以讲解一下bbox refinement么？

  #22 opened 2 years ago by flibbertigibbet-Y
  2

• box refine

  #24 opened 2 years ago by lina926
  1

• about reg_loss

  #26 opened 2 years ago by ycyinchao
  4

• What is the role of CenterPrior？

  #27 opened 2 years ago by ycyinchao
  2

• where is the code of the Probability of being a Negative Sample when IoU between 0.5 and 0.95?

  #25 opened 2 years ago by ycyinchao
  2

• Why not limit the regression range for each location of each FPN level?

  #17 opened 2 years ago by Icecream-blue-sky
  1

• What does the O(B, 0) in this equation mean?

  #7 opened 3 years ago by JinYuannn
  3

• The question about that the pos weight of each anchor for each instance is normalized by the sum of all pos weights within the candidate bag.

  #23 opened 2 years ago by MooresS
  4

• The multi-scale setting

  #20 opened 2 years ago by jshilong
  2

• some issues about code

  #19 opened 2 years ago by jacksonsc007
  3

• label assignment in overlap situation

  #18 opened 2 years ago by jacksonsc007
  2

• If the DW can be applied to softmax classifier?

  #15 opened 2 years ago by Icecream-blue-sky
  4

• Can't see the train loss in the log, the code directly do validation process

  #14 opened 3 years ago by Icecream-blue-sky
  0

• 运行出现错误

  #13 opened 3 years ago by sharp-xiang
  3

• Why the bbox_pred need to multiply with the stride?

  #12 opened 3 years ago by gyftsy
  3

• About `neg_metrics` -> `p_neg_weight`

  #11 opened 3 years ago by filaPro
  2

• About focal loss

  #10 opened 3 years ago by filaPro
  1

• Training in Custom Dataset

  #4 opened 3 years ago by BlueBlueFF
  1

• 原来博主是基粉

  #8 opened 3 years ago by dreamlychina
  0

• CUDA error: device-side assert triggered

  #9 opened 3 years ago by DDGRCF
  4

• Why to choose different exponentials during the calculation of pos weight ?

  #3 opened 3 years ago by soHardToHaveAName
  2

• Top-k method for selecting candidate bags

  #6 opened 3 years ago by soHardToHaveAName
  0

• ' DWHead is not in the models registry'"

  #5 opened 3 years ago by GF1447571253
  3

• how dose DW perform on Yolov5

  #2 opened 3 years ago by CangHaiQingYue
  1

• During training, the loss_cls_neg does not reduce, but the results appear good

  #1 opened 3 years ago by StarBurstStream0
  2